In [1]:
import cv2
import numpy as np
In [2]:
img = cv2.imread("./data/Dozen_0.bmp")
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
In [3]:
import matplotlib.pyplot
In [4]:
%matplotlib inline
matplotlib.pyplot.imshow(img)
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
In [5]:
gs = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
In [6]:
%matplotlib inline
matplotlib.pyplot.imshow(gs, "gray")
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
In [7]:
th = cv2.adaptiveThreshold(gs, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 63, 8)
th = ~th
In [8]:
%matplotlib inline
matplotlib.pyplot.imshow(th, "gray")
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
In [9]:
def get_threshold(gs):
th = cv2.adaptiveThreshold(gs, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 63, 8)
th = ~th
return th
def get_contours(gs, threshold=True):
if threshold:
th = cv2.adaptiveThreshold(gs, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 63, 8)
th = ~th
else:
th = gs.copy()
contours, hierarchy = cv2.findContours(th, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
return filter(lambda c: cv2.contourArea(c) > 3000.0, contours)
In [10]:
contours2 = get_contours(gs)
In [11]:
def get_inside(img, contours):
img2 = np.zeros(img.shape[:2], 'uint8')
cv2.drawContours(img2, contours, -1, 255, -1)
return img2
In [12]:
img2 = get_inside(img, contours2)
In [13]:
%matplotlib inline
matplotlib.pyplot.imshow(img2, 'gray')
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
In [14]:
def erode_contours(img, contours):
img2 = np.zeros(img.shape[:2], 'uint8')
cv2.drawContours(img2, contours, -1, 255, -1)
sz = cv2.arcLength(contours[0], True) / 20
kernel = np.ones((sz, sz),np.uint8)
erosion = cv2.erode(img2, kernel)
contours3, hierarchy3 = cv2.findContours(erosion, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
return contours3
In [15]:
contours3 = erode_contours(img, contours2)
In [16]:
def draw_countors(img, contours):
img3 = img.copy()
cv2.drawContours(img3, contours, -1, (0, 255, 0), 3)
return img3
In [17]:
img3 = draw_countors(img, contours3)
In [18]:
%matplotlib inline
matplotlib.pyplot.imshow(img3)
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
In [19]:
def get_ry(img, img2, r_thresh=-4, y_thresh=-4):
img_y = cv2.cvtColor(img, cv2.COLOR_RGB2YUV)
clahe = cv2.createCLAHE(clipLimit=10.0, tileGridSize=(32, 32))
channels = cv2.split(img_y)
channels[0] = clahe.apply(channels[0])
img_y = cv2.merge(channels, 3)
img_y = cv2.cvtColor(img_y, cv2.COLOR_YUV2RGB )
img_y = cv2.medianBlur(img_y, 13)
channels = cv2.split(img_y)
source = np.array(map(lambda x, y: map(lambda u, v: u - min(u, v), x, y), channels[0], channels[1]), dtype='uint8')
th_r = cv2.adaptiveThreshold(source, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 63, r_thresh)
th_r = cv2.bitwise_and(th_r, img2)
source = np.array(map(lambda x, y, z: map(lambda u, v, w: v - min(u / 2 + w / 2, v), x, y, z),
channels[0], channels[1], channels[2]), dtype='uint8')
th_y = cv2.adaptiveThreshold(source, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 63, y_thresh)
th_y = cv2.bitwise_and(th_y, img2)
th_y = cv2.morphologyEx(th_y, cv2.MORPH_OPEN, np.ones((5, 5),np.uint8))
th_y = cv2.morphologyEx(th_y, cv2.MORPH_CLOSE, np.ones((5, 5),np.uint8))
th_y = cv2.morphologyEx(th_y, cv2.MORPH_DILATE, np.ones((1, 1),np.uint8))
th_r = cv2.morphologyEx(th_r, cv2.MORPH_OPEN, np.ones((5, 5),np.uint8))
th_r = cv2.morphologyEx(th_r, cv2.MORPH_CLOSE, np.ones((5, 5),np.uint8))
th_r = cv2.morphologyEx(th_r, cv2.MORPH_DILATE, np.ones((1, 1),np.uint8))
contours, hierarchy = cv2.findContours(th_y.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
contours = filter(lambda c: cv2.contourArea(c) < 100.0, contours)
inside = get_inside(img, contours)
for i in range(len(img)):
for j in range(len(img[0])):
if inside[i][j]:
th_y[i][j] = 0
contours, hierarchy = cv2.findContours(th_r.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
contours = filter(lambda c: cv2.contourArea(c) < 100.0, contours)
inside = get_inside(img, contours)
for i in range(len(img)):
for j in range(len(img[0])):
if inside[i][j]:
th_r[i][j] = 0
return th_r, th_y
In [20]:
th_r, th_y = get_ry(img, img2)
In [21]:
%matplotlib inline
matplotlib.pyplot.imshow(img)
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
matplotlib.pyplot.imshow(th_r, 'gray')
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
matplotlib.pyplot.imshow(th_y, 'gray')
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
In [22]:
def count_labels(l, verbose=False):
sz = len(l)
start = 0
old = -1
labels = []
counts12 = []
counts = []
for i in range(sz):
if l[i] != 0:
start = i
break
i = start
for j in range(sz):
if l[i] == 0:
start = i
break
i =(i + 1) % sz
i = start
l2 = map(lambda x: x if x != 2 else 1, l)
for j in range(sz):
if l2[i] == old:
counts[-1] += 1
counts12[-1][l[i] - 1] += 1
else:
old = l2[i]
labels.append(l2[i])
counts12.append([0, 0])
counts12[-1][l[i] - 1] += 1
counts.append(1)
i = (i + 1) % sz
for i in range(len(labels)):
if (labels[i] == 1) and (3 * counts12[i][1] / 2 > counts12[i][0]):
labels[i] = 2
if verbose:
print labels, counts
cnt_sz = len(labels)
labels_out = cnt_sz * [0]
counts_1 = list(counts)
for i in filter(lambda j: labels[j] != 0, range(cnt_sz)):
counts_1[i] = -1
pos1 = np.argmax(counts_1)
max1 = np.max(counts_1)
labels_out[pos1] = 3
counts_1[pos1] = -1
pos2 = np.argmax(counts_1)
max2 = np.max(counts_1)
if max1 < 3 * max2 / 2:
labels_out[pos2] = 3
else:
labels_out[pos1] = 4
counts_1 = list(counts)
for i in filter(lambda j: labels[j] != 1, range(cnt_sz)):
counts_1[i] = -1
pos1 = np.argmax(counts_1)
max1 = np.max(counts_1)
labels_out[pos1] = 1
counts_1[pos1] = -1
pos2 = np.argmax(counts_1)
max2 = np.max(counts_1)
if 4 < max2:
labels_out[pos2] = 1
else:
labels_out[pos1] = 5
counts_1 = list(counts)
for i in filter(lambda j: labels[j] != 2, range(cnt_sz)):
counts_1[i] = -1
pos1 = np.argmax(counts_1)
max1 = np.max(counts_1)
labels_out[pos1] = 2
counts_1[pos1] = -1
pos2 = np.argmax(counts_1)
max2 = np.max(counts_1)
if 4 < max2:
labels_out[pos2] = 2
else:
labels_out[pos1] = 6
labels = []
for i in labels_out:
if i == 4:
labels.append(3)
labels.append(3)
elif i == 5:
labels.append(1)
labels.append(1)
elif i == 6:
labels.append(2)
labels.append(2)
else:
labels.append(i)
return filter(lambda x: x != 0, labels)
In [23]:
def contour_label(cnt, th_y, th_r, verbose=False):
pixels = map(lambda e: (th_y[e[0][1], e[0][0]], th_r[e[0][1], e[0][0]]), cnt)
color_labels = map(lambda p: 2 if p[0] else 1 if p[1] else 0, pixels)
if verbose:
print color_labels
return count_labels(color_labels, verbose)
1. Определить количество фишек на изображении. Входом является файл типа Group_*.bmp.¶
In [24]:
def item_count(filename):
img = cv2.imread(filename)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
gs = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
contours = get_contours(gs)
img3 = draw_countors(img, contours)
matplotlib.pyplot.imshow(img3)
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
return len(contours)
In [25]:
%matplotlib inline
print item_count("./data/Group_1.bmp"), 'items'
In [26]:
%matplotlib inline
print item_count("./data/Group_2.bmp"), 'items'
In [27]:
%matplotlib inline
print item_count("./data/Group_3.bmp"), 'items'
In [28]:
%matplotlib inline
print item_count("./data/Group_4.bmp"), 'items'
In [29]:
%matplotlib inline
print item_count("./data/Group_5.bmp"), 'items'
In [30]:
%matplotlib inline
print item_count("./data/Group_6.bmp"), 'items'
2. Определить тип и цвет линий на фишке – короткая дуга большой кривизны, длинная дуга малой кривизны, прямолинейный сегмент. Вход – файл типа Single_*.bmp.¶
In [31]:
def segment_length(cl, idx):
l = []
for i in range(len(cl)):
if cl[i] == idx:
l.append(i)
a, b = l[0], l[1]
return min(abs(a - b), abs(len(cl) + a - b), abs(len(cl) + b - a))
def semgent_type(filename):
img = cv2.imread(filename)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
gs = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
contours = get_contours(gs)
img3 = draw_countors(img, contours)
matplotlib.pyplot.imshow(img3)
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
img2 = get_inside(img, contours)
contours3 = erode_contours(img, contours)
th_r, th_y = get_ry(img, img2)
cl = contour_label(contours3[0], th_y, th_r)
return segment_length(cl, 1), segment_length(cl, 2), segment_length(cl, 3)
In [32]:
r_len, y_len, b_len = semgent_type("./data/Single_0.bmp")
print "Red length =", r_len
print "Yellow length =", y_len
print "Blue length =", b_len
In [33]:
r_len, y_len, b_len = semgent_type("./data/Single_1.bmp")
print "Red length =", r_len
print "Yellow length =", y_len
print "Blue length =", b_len
In [34]:
r_len, y_len, b_len = semgent_type("./data/Single_2.bmp")
print "Red length =", r_len
print "Yellow length =", y_len
print "Blue length =", b_len
In [35]:
r_len, y_len, b_len = semgent_type("./data/Single_3.bmp")
print "Red length =", r_len
print "Yellow length =", y_len
print "Blue length =", b_len
In [36]:
r_len, y_len, b_len = semgent_type("./data/Single_4.bmp")
print "Red length =", r_len
print "Yellow length =", y_len
print "Blue length =", b_len
In [37]:
r_len, y_len, b_len = semgent_type("./data/Single_5.bmp")
print "Red length =", r_len
print "Yellow length =", y_len
print "Blue length =", b_len
In [38]:
r_len, y_len, b_len = semgent_type("./data/Single_6.bmp")
print "Red length =", r_len
print "Yellow length =", y_len
print "Blue length =", b_len
In [39]:
r_len, y_len, b_len = semgent_type("./data/Single_7.bmp")
print "Red length =", r_len
print "Yellow length =", y_len
print "Blue length =", b_len
In [40]:
r_len, y_len, b_len = semgent_type("./data/Single_8.bmp")
print "Red length =", r_len
print "Yellow length =", y_len
print "Blue length =", b_len
In [41]:
r_len, y_len, b_len = semgent_type("./data/Single_9.bmp")
print "Red length =", r_len
print "Yellow length =", y_len
print "Blue length =", b_len
3. Определить номер фишки. Вход – файл типа Single_*.bmp.¶
In [42]:
def draw_text(img, s, coord):
cv2.putText(img, s, (coord[0], coord[1]),
cv2.FONT_HERSHEY_COMPLEX_SMALL, 4.0, (255, 255, 255), 5, cv2.CV_AA);
def contour_cmp(a, b):
if a[0][0][0] < b[0][0][0]:
return -1
if a[0][0][0] > b[0][0][0]:
return 1
return 0
def cl_to_str(cl):
s = ""
for i in cl:
s += str(i)
return s
def get_database(filename):
img = cv2.imread(filename)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
gs = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
contours = get_contours(gs)
img3 = draw_countors(img, contours)
img2 = get_inside(img, contours)
contours3 = sorted(erode_contours(img, contours), cmp=contour_cmp)
for i in range(len(contours3)):
draw_text(img3, str(i + 1), contours3[i][0][0])
th_r, th_y = get_ry(img, img2)
return map(lambda x: (cl_to_str(contour_label(contours3[x], th_y, th_r)), x + 1), range(len(contours3)))
In [43]:
db = get_database("./data/Dozen_0.bmp")
In [44]:
print db
In [45]:
def item_number(filename, db, verbose=False):
img = cv2.imread(filename)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
gs = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
contours = get_contours(gs)
img3 = draw_countors(img, contours)
img2 = get_inside(img, contours)
contours3 = erode_contours(img, contours)
th_r, th_y = get_ry(img, img2)
if verbose:
matplotlib.pyplot.imshow(th_r, 'gray')
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
matplotlib.pyplot.imshow(th_y, 'gray')
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
for i in range(len(contours3)):
cl = contour_label(contours3[i], th_y, th_r, verbose=verbose)
p = cl_to_str(cl)
for j in db:
s = j[0] + j[0]
if s.find(p) != -1:
item_num = j[1]
break
if verbose:
print(item_num)
draw_text(img3, str(item_num), contours3[i][0][0])
matplotlib.pyplot.imshow(img3)
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
In [46]:
item_number("./data/Single_0.bmp", db)
In [47]:
item_number("./data/Single_1.bmp", db)
In [48]:
item_number("./data/Single_2.bmp", db)
In [49]:
item_number("./data/Single_3.bmp", db)
In [50]:
item_number("./data/Single_4.bmp", db)
In [51]:
item_number("./data/Single_5.bmp", db)
In [52]:
item_number("./data/Single_6.bmp", db)
In [53]:
item_number("./data/Single_7.bmp", db)
In [54]:
item_number("./data/Single_8.bmp", db)
In [55]:
item_number("./data/Single_9.bmp", db)
4. Определить расположение и номера всех фишек в кадре. Вход – файл типа Group_*.bmp.¶
In [56]:
item_number("./data/Group_1.bmp", db)
In [57]:
item_number("./data/Group_2.bmp", db)
In [58]:
item_number("./data/Group_3.bmp", db)
In [59]:
item_number("./data/Group_4.bmp", db)
In [60]:
item_number("./data/Group_5.bmp", db)
In [61]:
item_number("./data/Group_6.bmp", db)
5. Определить последовательность обхода фишек в мозаике вдоль замкнутого маршрута. Вход – файл типа Path_*.bmp.¶
In [62]:
img = cv2.imread("./data/Path_5.bmp")
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
gs = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY) #оттенки серого
In [63]:
contours2 = get_contours(gs) #контур
In [64]:
img2 = get_inside(img, contours2) # бинаризация
In [65]:
%matplotlib inline
matplotlib.pyplot.imshow(img2, 'gray')
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
In [66]:
selem = np.ones((9, 9))
dilated = cv2.morphologyEx(img2, cv2.MORPH_CLOSE, selem)
In [67]:
def show_binar(img):
%matplotlib inline
matplotlib.pyplot.imshow(img, 'gray')
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
In [68]:
show_binar(dilated)
In [69]:
def show_col(img):
%matplotlib inline
matplotlib.pyplot.imshow(img)
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
In [70]:
contours2 = get_contours(img2, False)
img3 = draw_countors(img, contours2)
show_col(img3)
In [71]:
contours2 = get_contours(dilated, False)
img3 = draw_countors(img, contours2)
show_col(img3)
In [72]:
ap = cv2.approxPolyDP(contours2[0], cv2.arcLength(contours2[0], True) / 100, True)
img4 = draw_countors(img, ap)
show_col(img4)
In [73]:
cv2.arcLength(contours2[0], True)
Out[73]:
In [74]:
ap2 = map(lambda x: x[0], ap)
In [75]:
def get_angle(a, b):
an = a / np.linalg.norm(a)
bn = b / np.linalg.norm(b)
return np.dot(an, bn)
In [76]:
result = []
for i in range(len(ap2)):
p1 = ap2[i]
p2 = ap2[(i + 1) % len(ap2)]
p3 = ap2[(i + 2) % len(ap2)]
v1 = p1 - p2
v2 = p3 - p2
result.append(get_angle(v1, v2))
In [77]:
def rotation_matrix(angle):
return np.array([[np.cos(angle), np.sin(angle)], [-np.sin(angle), np.cos(angle)]])
def add_points(p1, p2, new_points):
v = p2 - p1
m1 = rotation_matrix(+ 60.0 / 180.0 * np.pi)
m2 = rotation_matrix(- 60.0 / 180.0 * np.pi)
new_points.append(p2 + np.dot(m1, v))
new_points.append(p2 + np.dot(m2, v))
def points_to_contour(points):
return np.array(map(lambda x: [x], points), 'int32')
def hagrid(points, polygon):
new_points = []
slen = 0.0
for i in range(len(points)):
p1 = points[i]
p2 = points[(i + 1) % len(points)]
new_points.append(p1)
new_points.append(p2)
slen += np.linalg.norm(p2 - p1)
add_points(p1, p2, new_points)
add_points(p2, p1, new_points)
slen /= len(points)
result = []
for i in points:
result.append(i)
for i in new_points:
ok = False
for j in range(len(points)):
if np.linalg.norm(i - points[j]) < slen * 0.5:
ok = True
break
if ok:
continue
ok = False
for j in range(len(result)):
if np.linalg.norm(i - result[j]) < slen * 0.5:
ok = True
result[j] = (0.5 * result[j] + 0.5 * i)
break
if not ok:
result.append(i)
return filter(lambda x: cv2.pointPolygonTest(polygon, (x[0], x[1]), True) > - slen * 0.5, result)
In [78]:
ap3 = hagrid(ap2, points_to_contour(ap2))
ap3 = hagrid(ap3, points_to_contour(ap2))
img5 = draw_countors(img, points_to_contour(ap3))
show_col(img5)
In [79]:
def build_graph(ap3):
g = []
for i in range(len(ap3)):
vl = []
min_dist = -1
for j in range(len(ap3)):
if i == j:
continue
cur_dist = np.linalg.norm(ap3[i] - ap3[j])
if (min_dist == -1) or (cur_dist < min_dist):
min_dist = cur_dist
for j in range(len(ap3)):
if i == j:
continue
cur_dist = np.linalg.norm(ap3[i] - ap3[j])
if abs(cur_dist - min_dist) < min_dist * 0.5:
vl.append(j)
g.append(vl)
for i in range(len(g)):
for j in g[i]:
found = False
for k in g[j]:
if k == i:
found = True
break
if not found:
g[j].append(i)
return g
In [80]:
img6 = img.copy()
for i in range(len(ap3)):
draw_text(img6, str(i), (int(ap3[i][0]), int(ap3[i][1])))
show_col(img6)
In [81]:
g = build_graph(ap3)
In [82]:
g
Out[82]:
In [83]:
def dfs(g, v, p, st, vp, path=[], depth=0):
if depth == 6:
if v == st:
vp.append(path)
return
for u in g[v]:
if p != u:
dfs(g, u, v, st, vp, path + [u], depth + 1)
def clockwise(ap, path):
p1 = ap[path[0]]
p2 = ap[path[1]]
p3 = ap[path[2]]
v1 = p2 - p1
v2 = p3 - p2
return np.cross(v1, v2)
In [84]:
def get_paths(g, ap3, v):
vp = []
dfs(g, v, -1, v, vp)
return filter(lambda x: clockwise(ap3, x) < 0, vp)
In [85]:
def path_similiar(p1, p2):
st = 0
for i in range(len(p2)):
if p2[i] == p1[0]:
st = i
break
for i in range(len(p1)):
if p1[i] != p2[st]:
return False
st = (st + 1) % len(p2)
return True
def add_path(paths, p):
for i in range(len(p)):
for j in range(len(p)):
if i == j:
continue
if p[i] == p[j]:
return
ok = False
for cur in paths:
if path_similiar(cur, p):
ok = True
break
if not ok:
paths.append(p)
In [86]:
def get_all_paths(g, ap3):
paths = []
for i in range(len(g)):
p = get_paths(g, ap3, i)
for j in p:
add_path(paths, j)
return paths
In [87]:
paths = get_all_paths(g, ap3)
In [88]:
def contour_center(cnt):
moments = cv2.moments(cnt)
return int(moments['m10'] / moments['m00']), int(moments['m01'] / moments['m00'])
def erode_item_contours(img, item_contours):
result = []
for i in range(len(item_contours)):
img2 = np.zeros(img.shape[:2], 'uint8')
cv2.drawContours(img2, [item_contours[i]], -1, 255, -1)
sz = cv2.arcLength(item_contours[0], True) / 20
kernel = np.ones((sz, sz), np.uint8)
erosion = cv2.erode(img2, kernel)
contours3, hierarchy3 = cv2.findContours(erosion, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
if len(contours3) > 0:
result.append(contours3[0])
return result
return np.array(result, 'int32')
def item_path(img, global_contours, item_contours, paths, ic_indices, db, verbose=False, r_thresh=-4, y_thresh=-4):
gs = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
img3 = draw_countors(img, item_contours)
img2 = get_inside(img, global_contours)
contours3 = erode_item_contours(img, item_contours)
th_r, th_y = get_ry(img, img2, r_thresh=r_thresh, y_thresh=y_thresh)
if verbose:
matplotlib.pyplot.imshow(th_r, 'gray')
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
matplotlib.pyplot.imshow(th_y, 'gray')
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
img_c = draw_countors(img, contours3)
matplotlib.pyplot.imshow(img_c)
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
item_numbers = len(contours3) * [0]
for i in range(len(contours3)):
cl = contour_label(contours3[i], th_y, th_r, verbose)
if verbose:
print cl
p = cl_to_str(cl)
item_num = -1
for j in db:
s = j[0] + j[0]
if s.find(p) != -1:
item_num = j[1]
break
item_numbers[i] = item_num
item_set = set(range(1, 11))
for i in item_numbers:
if i in item_set:
item_set.remove(i)
for i in range(len(item_numbers)):
if item_numbers[i] == -1:
for k in item_set:
item_set.remove(k)
item_numbers[i] = k
break
for i in range(len(contours3)):
if verbose:
print item_numbers[i]
draw_text(img3, str(item_numbers[i]), contour_center(item_contours[i]))
matplotlib.pyplot.imshow(img3)
matplotlib.pyplot.axis('off')
matplotlib.pyplot.show()
return item_numbers
In [89]:
def path_to_contour(p, ap3):
return np.array(map(lambda i: [ap3[i]], p), 'int32')
def paths_to_contours(p, ap3):
return np.array(map(lambda cur: path_to_contour(cur, ap3), p), 'int32')
In [90]:
def is_hole(img, cnt):
th = get_threshold(img)
img2 = np.zeros(img.shape[:2], 'uint8')
cv2.drawContours(img2, [cnt], -1, 255, -1)
cnt = 0.0
all = 1.0
for i in range(len(img2)):
for j in range(len(img2[i])):
if img2[i][j]:
all += 1.0
if th[i][j]:
cnt += 1.0
return cnt / all < 0.15
In [91]:
paths
Out[91]:
In [92]:
item_contours = paths_to_contours(paths, ap3)
ic_indices = range(len(paths))
r = filter(lambda cnt: not is_hole(gs, cnt[0]), zip(item_contours, ic_indices))
In [93]:
item_contours, ic_indices = zip(*r)
In [94]:
inn = item_path(img, contours2, item_contours, paths, ic_indices, db)
In [95]:
for i in inn:
print i,
In [96]:
def build_item_graph(paths, ic_indices):
fpaths = map(lambda x: paths[x], ic_indices)
result = []
for i in range(len(fpaths)):
tmp = []
for j in range(6):
tmp.append(-1)
result.append(tmp)
for i in range(len(result)):
old = fpaths[i][-1]
for j in range(6):
cur = fpaths[i][j]
for k in range(len(result)):
if i == k:
continue
oldk = fpaths[k][-1]
for l in range(6):
curk = fpaths[k][l]
if (curk == old) and (cur == oldk):
result[i][j] = k
oldk = curk
old = cur
return result
In [97]:
item_graph = build_item_graph(paths, ic_indices)
In [98]:
for i in item_graph:
for j in i:
print j,
print
In [99]:
for i in db:
for j in i[0]:
print j,
print
Компиляция СИ-ПРОГРАММЫ¶
In [100]:
!cd ./c && ./compile.sh
Запись входных данных¶
In [101]:
tmpfile = open('./c/tmpfile', 'w')
In [102]:
for i in db:
for j in i[0]:
tmpfile.write(str(j) + ' ')
tmpfile.write('\n')
In [103]:
tmpfile.write(str(len(inn)) + '\n')
In [104]:
for i in inn:
tmpfile.write(str(i) + ' ')
tmpfile.write('\n')
In [105]:
for i in item_graph:
for j in i:
tmpfile.write(str(j) + ' ')
tmpfile.write('\n')
In [106]:
tmpfile.close()
Запуск СИ-ПРОГРАММЫ¶
In [107]:
!cd ./c && ./main <tmpfile >tmpfile_out
Считывание выходных данных СИ-ПРОГРАММЫ¶
In [108]:
tmpfile = open('./c/tmpfile_out', 'r')
In [109]:
output = map(lambda x: int(x), tmpfile.readline()[:-2].split(' '))
In [110]:
print output
Собираем всё в кучу¶
In [111]:
def detect_path(filename):
img = cv2.imread(filename)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
gs = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
contours2 = get_contours(gs) #контур
img2 = get_inside(img, contours2) # бинаризация
selem = np.ones((9, 9))
dilated = cv2.morphologyEx(img2, cv2.MORPH_CLOSE, selem)
contours2 = get_contours(img2, False)
img3 = draw_countors(img, contours2)
contours2 = get_contours(dilated, False)
img3 = draw_countors(img, contours2)
ap = cv2.approxPolyDP(contours2[0], cv2.arcLength(contours2[0], True) / 100, True)
img4 = draw_countors(img, ap)
ap2 = map(lambda x: x[0], ap)
result = []
for i in range(len(ap2)):
p1 = ap2[i]
p2 = ap2[(i + 1) % len(ap2)]
p3 = ap2[(i + 2) % len(ap2)]
v1 = p1 - p2
v2 = p3 - p2
result.append(get_angle(v1, v2))
ap3 = hagrid(ap2, points_to_contour(ap2))
ap3 = hagrid(ap3, points_to_contour(ap2))
img5 = draw_countors(img, points_to_contour(ap3))
g = build_graph(ap3)
paths = get_all_paths(g, ap3)
item_contours = paths_to_contours(paths, ap3)
ic_indices = range(len(paths))
r = filter(lambda cnt: not is_hole(gs, cnt[0]), zip(item_contours, ic_indices))
item_contours, ic_indices = zip(*r)
inn = item_path(img, contours2, item_contours, paths, ic_indices, db)
item_graph = build_item_graph(paths, ic_indices)
tmpfile = open('./c/tmpfile', 'w')
for i in db:
for j in i[0]:
tmpfile.write(str(j) + ' ')
tmpfile.write('\n')
tmpfile.write(str(len(inn)) + '\n')
for i in inn:
tmpfile.write(str(i) + ' ')
tmpfile.write('\n')
for i in item_graph:
for j in i:
tmpfile.write(str(j) + ' ')
tmpfile.write('\n')
tmpfile.close()
!cd ./c && ./main <tmpfile >tmpfile_out
tmpfile = open('./c/tmpfile_out', 'r')
output = map(lambda x: int(x), tmpfile.readline()[:-2].split(' '))
tmpfile.close()
return output
In [112]:
result = detect_path("./data/Path_4.bmp")
print "Path:",
for i in result:
print i,
In [113]:
result = detect_path("./data/Path_5.bmp")
print "Path:",
for i in result:
print i,
In [114]:
result = detect_path("./data/Path_7_1.bmp")
print "Path:",
for i in result:
print i,
In [115]:
result = detect_path("./data/Path_9.bmp")
print "Path:",
for i in result:
print i,
Запуск тестов¶
Количество фишек¶
In [116]:
%matplotlib inline
print item_count("./data/Group_1.bmp"), 'items'
Определение типов сегментов¶
In [117]:
r_len, y_len, b_len = semgent_type("./data/Single_0.bmp")
print "Red length =", r_len
print "Yellow length =", y_len
print "Blue length =", b_len
Определение номеров фишек¶
In [118]:
item_number("./data/Group_1.bmp", db)
Нахождение пути¶
In [119]:
result = detect_path("./data/Path_4.bmp")
print "Path:",
for i in result:
print i,